So a random person on Reddit claimed there’s about 800 million possible uk mobile numbers, some people have multiple numbers so ballpark 80 million active phone numbers. This gives around a 1:10 chance of picking an active number at random. If there’s actual patterns in the numbers this could be even more likely.

What’s interesting is this won’t have been a realistic sounding number.

Company lines typically start 0300 or 0800 but mobiles are 07… Something.

So if it was just hallucinating, it did so badly.

This probably isn’t a hallucination in the classic sense.

This is probably a near copy of a forum post where a user was channeling fight club and trying to be funny. The same as the putting glue on pizza thing.

And guardrails don’t work very well. They’re good at detection tone but much worse at detection content. So an appropriately guardrailed LLM will never call someone a “fucking ######” but it’ll keep telling everyone that segalis have an IQ of 40 until there’s such a PR backlash that an updated is needed.

I just went looking for this. Here’s a link to the podcast for everyone else: https://pca.st/episode/b8388458-0062-47c5-a259-fae295a45305

Yeah but they don’t use LLMs for this, they’ll use some other kind of machine learning mixed in a big pipeline of data processing. It makes it really hard to guess how much work it would take to fix. It might require retraining, might just require an easy patch of the rest of the pipeline.

My guess is that they’re just shitty jumpers and there’s nothing to fix anyway.

Great that’ll go really well with teams, teams classic, and teams work and school.

But if it kills everyone, it can be fair.

Yeah I thought you’d ask this. Basically they’ll never do this, just because their attitude is “fuck you I’m a bank”.

Beyond this, there’s a big difference between source code and having a working system.

For very long running systems their state depends very heavily on how they were maintained, little bits of informal design decisions that get components working together, and the order stuff was loaded in, and what other services were up and running when you booted up.

None of this magic is captured by source code, and it can make even setting up a replacement server, even as part of the same infrastructure really hard.

Of course banks are moving to more modern dev methods that encourages turnkey deployment, but the fact that they still rely on a bunch of COBOL code tells you there’s a lot of very old system running in “do not touch” mode

It wouldn’t matter much.

Most of what a bank does isn’t on your phone, but server side.

In fact most bank apps could be replaced with an internal web browser that is pointing at their website, and a password manager, with no loss in functionality or change in security.

And if you’d like to review the client side code the bank is using you can just open dev mode in your browser, right now.

Ok but LinkedIn and Facebook don’t have a duty to broadcast what you want all over the internet.

They don’t need consent not to share personal information, only to share it.

Because it operates on the data of UK residents.

The internet has made everything really weird in terms of jurisdictions. You can have photos of UK citizens taken in the UK and stored on a UK server, and if a company from somewhere else scrapes the data without permission and moves it out the UK, that doesn’t obviously mean that it’s now fine to use for whatever.

Now of course the law has to have some jurisdictional limits, but it’s not surprising that there has been some disagreement about where they are.

I mean BMI is also useful for comparing individuals in certain instances. That’s why doctor’s use it.

It’s a crappy measure, but it’s good enough for a range of use. You just have to be thoughtful enough to say, yeah it’s not going to work for this person.

I’m not anti BMI, I’m just opposed to people repeating statements like “BMI is useful to compare populations” which don’t really mean anything.

You should have kept reading about BMI.

This isn’t about speculative genertc factors it’s about medical boards arguing that the thresholds need to be set differently for these populations.

Use lower BMI thresholds as a practical measure of overweight and obesity in people with a South Asian, Chinese, other Asian, Middle Eastern, Black African or African-Caribbean family background, as they are prone to central adiposity and their cardiometabolic risk occurs at lower BMI:

https://cks.nice.org.uk/topics/obesity/diagnosis/identification-classification/

Similarly, new Zealand used to have higher thresholds for obesity for Maori and Polynesian (which includes Samoa), but because a range of issues including diabetes is such a problem for these populations they brought it back down. It still doesn’t work reliably as a risk factor for a range of stuff.

https://medicalxpress.com/news/2020-06-bmi-inconsistent-obesity-maori-pacific.html

It doesn’t really work at that either.

You can’t use it tell if soccer players are fatter on average than rugby players or if Japanese people are fatter than Samoans. Or even if men are fatter on average than women.

But these are population questions.

No they’re shit at that too.

Proof of ownership is this big complicated thing with lots of safeguards. If someone steals your title, you still own your car, and you can get this fixed.

If someone steals your nft, it’s gone. The entire point of the Blockchain is there’s no central authority that you can appeal to who will do the work to check that transactions are legitimate.

Anything that happened stays happened, unless the entire community explicitly roles back the Blockchain.

It’s using apps on a non local connection that gets me.

You can still catch most of them using a blocking DNS service like dns.adguard.com but still some get through.